Transformations¶

Transformations of the time series intended to be used in a similar fashion to torchvision.

class transforms.Compose(transforms)¶

Composes several transforms together.

List of transforms must currently begin with ToTensor and end with Target.

Args:

transforms (list of Transform objects): list of transforms to compose.

Example:

>>> transforms.Compose([
>>>     transforms.ToTensor(),
>>>     transforms.LogTransform(targets=[0], offset=1.0),
>>>     transforms.Target(targets=[0]),
>>> ])

class transforms.LogTransform(targets=None, offset=0.0)¶

Natural logarithm of target covariate + offset.

\[y_i = log_e ( x_i + \mbox{offset} )\]

Args:

offset (float): amount to add before taking the natural logarithm
targets (list): list of indices to transform.

Example:

>>> transforms.LogTransform(targets=[0], offset=1.0)

class transforms.RemoveLast(targets=None)¶

Subtract final point in lookback window from all points in example.

Args:

targets (list): list of indices to transform.

Example:

>>> transforms.RemoveLast(targets=[0])

class transforms.Standardize(targets=None)¶

Subtract the mean and divide by the standard deviation from the lookback.

Args:

targets (list): list of indices to transform.

Example:

>>> transforms.Standardize(targets=[0])

class transforms.Target(targets)¶

Retain only target indices for output.

Args:

targets (list): list of indices to retain.

Example:

>>> transforms.Target(targets=[0])

class transforms.ToTensor(device='cpu')¶

Convert numpy.ndarrays to tensor.

Args:

device (str): device on which to load the tensor.

Example:

>>> transforms.ToTensor(device='cpu')